We had some templates fail to deploy on ci.jenkins.io, upon further investigation I discovered that the reason the ARM deploy had failed is because the resource group was exceeding a quota.
It turns out, Azure has a *lot* more VMs spun up than our Jenkins currently has attached as agents.
There are *six* VMs running in Azure and basically none are registered as agents in Jenkins. I don't know how long they all have been running
Not sure what diagnostics could help, i'm going to delete these excess VMs so I can stop burning money.